Hyperpolarized 129Xe Time-of-Flight MR Imaging of Perfusion and Brain Function

Perfusion measurements can provide vital information about the homeostasis of an organ and can therefore be used as biomarkers to diagnose a variety of cardiovascular, renal, and neurological diseases. Currently, the most common techniques to measure perfusion are 15O positron emission tomography (PET), xenon-enhanced computed tomography (CT), single photon emission computed tomography (SPECT), dynamic contrast enhanced (DCE) MRI, and arterial spin labeling (ASL) MRI. Here, we show how regional perfusion can be quantitively measured with magnetic resonance imaging (MRI) using time-resolved depolarization of hyperpolarized (HP) xenon-129 (129Xe), and the application of this approach to detect changes in cerebral blood flow (CBF) due to a hemodynamic response in response to brain stimuli. The investigated HP 129Xe Time-of-Flight (TOF) technique produced perfusion images with an average signal-to-noise ratio (SNR) of 10.35. Furthermore, to our knowledge, the first hemodynamic response (HDR) map was acquired in healthy volunteers using the HP 129Xe TOF imaging. Responses to visual and motor stimuli were observed. The acquired HP TOF HDR maps correlated well with traditional proton blood oxygenation level-dependent functional MRI. Overall, this study expands the field of HP MRI with a novel dynamic imaging technique suitable for rapid and quantitative perfusion imaging.


Introduction
Perfusion measurements can provide vital information about the homeostasis of an organ [1] and can therefore be used as biomarkers to diagnose cardiovascular [2,3], renal [4], and neurological [5][6][7] diseases. Currently, the most commonly used techniques to measure perfusion are 15 O positron emission tomography (PET) [8][9][10], xenon-enhanced computed tomography (CT) [11,12], single photon emission computed tomography (SPECT) [3,8,13], and arterial spin labeling (ASL) magnetic resonance imaging (MRI) [1,[14][15][16][17][18]. In addition, dynamic susceptibility contrast (DSC) and dynamic contrast enhanced (DCE) MRI are frequently used for perfusion imaging [19][20][21]. Although these techniques are well-established, each has some serious drawbacks. CT requires high-dose ionizing radiation, PET and SPECT rely on injection of radioactive contrast agents and the acquired PET images are of low resolution, and the signal-to-noise ratio (SNR) and contrast of ASL images is low. DSC and DCE MRI require injection of contrast agents, most of which contain gadolinium, which was recently associated with a certain amount of toxicity [22,23]. In addition, the most commonly used gadolinium-based contrast agents are uncapable of crossing the blood-brain barrier [24][25][26], which makes cerebral perfusion imaging with these agents more challenging.
Hyperpolarized (HP) xenon-129 ( 129 Xe) MRI is a powerful MRI approach used mainly for lung imaging and for the study of lung disorders [27,28]. A hyperpolarized (HP) metastable state is produced by spin exchange optical pumping [29][30][31], and is characterized by up to 10 5 larger longitudinal magnetization compared to thermal polarization [27]. Therefore, the magnetic resonance signal of HP nuclei can be up to 10 5 times stronger than at thermal equilibrium. Due to the ability of HP 129 Xe to dissolve in blood and travel to highly perfused organs, HP 129 Xe MRI was recently used to study the brain [6,[32][33][34] and kidneys [35]. The last achievements in the field of HP 129 Xe MRI allowed the investigation of cerebral perfusion changes associated with Alzheimer's disease [34] and stroke [6].
Since HP 129 Xe dissolves in the blood [36] and has extreme sensitivity to chemical environments [27], it can be used as a contrast agent to study blood flow and conduct perfusion measurements in tissues. Since the HP state is a nonequilibrium, metastable state, the longitudinal magnetization is not restored by spin-lattice relaxation once a radiofrequency (RF) pulse irradiates the nuclei. After irradiation of a volume element containing HP 129 Xe dissolved in tissue or blood with a 90 • RF pulse, the HP state is completely destroyed and the dissolved HP 129 Xe does not produce any significant amount of signal. If there is continuous flow into the volume of dissolved 129 Xe, and if the MR measurement is conducted following a prescribed time delay (time-of-flight (TOF) time), the MR signal is determined mainly by the amount of 129 Xe washed into the selected volume. We hypothesize that it is possible to measure the blood flow and tissue perfusion quantitatively by creating a dynamic imaging technique which measures the 129 Xe signal evolution with change in TOF. The proposed perfusion imaging technique has the potential to open a new pathway for imaging and diagnostics of perfusion-related diseases.
In this proof-of-concept study, we develop a novel HP 129 Xe time-of-flight perfusion imaging pulse sequence and evaluate the performance of this technique in vitro and in vivo. The first quantitative HP cerebral perfusion images were acquired using the proposed technique in healthy volunteers. Furthermore, the HP 129 Xe TOF technique was used for hemodynamic response detection and was corroborated by conventional proton ( 1 H) blood oxygenation level-dependent (BOLD) fMRI.

Pulse Sequence Design
The Chemical Shift Saturation Recovery (CSSR) pulse sequence is typically used to study the time course of hyperpolarized (HP) 129 Xe gas exchange in the lungs [37,38]. We modified the CSSR sequence for cerebral perfusion imaging using HP 129 Xe to create the 129 Xe TOF used in this study ( Figure 1). The first 90 • block pulse has a time duration of 0.5 ms, which yields a bandwidth (BW) of 2 kHz (56.5 ppm). This pulse is broad enough to saturate the HP 129 Xe in all brain tissues and the Xe dissolved in blood. After termination of the saturation pulse, the unsaturated 129 Xe from the lungs flows into the brain during time τ. The image pulse sequence is then initiated. Here, the broad-band excitation pulse of spredrex shape was used for imaging. The image was acquired using a Gradient Echo (GRE) pulse sequence with a Cartesian read-out. The TOF pulse sequence was repeated three times with different recovery times τ (Figure 1). To detect hemodynamic response to a visual stimulus, we used the following delay times: τ 0 = 1 s, τ 1 = 6.5 s, τ 2 = 8 s. To detect hemodynamic response to a motor task, the TOF delays were equal to 2.5 s, 6.8 s, and 7.1 s.

In Vitro Flow Measurements
The flow phantom used in this study is shown in Figure 2a. The 3.175 mm tube was connected to the syringe pump (Kent Scientific Co, Torrington, CT, USA). A 30 mL syringe was filled with saline and loaded into the pump. One liter of natural abundant 129 Xe (~26%) was polarized up to 52% using the Xemed xenon polarizer and dispensed into a TedLar bag. A TedLar bag and a saline tube were connected to the membrane contactor (3M Liqui-Cel MM-0.5 × 1 Series) to mix HP Xe with saline. The outlet of the membrane contactor was connected to the 3.175 mm tube and the end of the tube was placed into the custom-built quadrature coil tuned to the Xe resonance frequency at 3T (35.33 MHz). A saline flow rate was set to 5 mL/min, 6 mL/min, 7 mL/min, and 10 mL/min.
The 129 Xe TOF GRE imaging was conducted using 20 TOF recovery times that varied from 200 ms to 2000 ms, with a step of 100 ms. The following parameters were used for imaging: TR/TE = 104.16 ms/1.22 ms, FA = 12.5°, FOV = 100 × 100 mm 2 .

In Vivo Cerebral Perfusion Measurements in Healthy Volunteers
This research study was approved (Jan. 28, 2019) by the research ethics boards (REB) at Lakehead University (LU) and the Thunder Bay Regional Health Sciences Centre (TBRHSC) (REB file number RP-307) and conducted in accordance with the Tri-Council Policy Statement-2 (TCPS-2). Nine healthy volunteers between the ages of 21 and 77 were recruited from the community for this study. All participants signed an informed consent form. All participants were cognitively normal and consented to the data obtained from them being disseminated in this report.

1 H Magnetic Resonance Imaging
Participants were placed into a dual tuned 1 H/ 129 Xe head coil in a Philips Achieva 3T clinical MRI scanner. T2-weighted 1 H MRI were acquired using a turbo-spin echo (TSE) sequence with a Cartesian readout. The high-resolution anatomical proton images were acquired in axial and sagittal views. The sagittal images were acquired using a field-of-view equal to 250 × 250 × 46 mm 3 and a voxel size of 0.98 × 1.04 × 6 mm 3 . The repetition time was set to 3 s and the echo time was equal to 80 s. The flip angle of excitation radiofrequency pulse was equal to 90°. Five slices were acquired, separated by 6 mm gaps. The axial scans were acquired using similar parameters. The only difference was in voxel size, which was equal to 0.98 × 0.99 × 6 mm 3 in the axial scans.

129 Xe Perfusion Mapping Using Time of Flight (TOF)
Enriched 129 Xe (83%) was polarized to ~50% using a Xemed xenon polarizer and dispensed into 1 L Tedlar bags. The participant breathed in the gas and held their breath for 20 s. The dynamic 129 Xe TOF imaging pulse sequence with a Cartesian gradient echo readout was initiated simultaneously with a breath-hold. Three dynamic projection brain images were acquired during the breath-hold. The TOF recovery delays were equal to 2.5 s, 6.7 s, and 7.1 s. The field of view of gradient echo pulse

In Vitro Flow Measurements
The flow phantom used in this study is shown in Figure 2a. The 3.175 mm tube was connected to the syringe pump (Kent Scientific Co, Torrington, CT, USA). A 30 mL syringe was filled with saline and loaded into the pump. One liter of natural abundant 129 Xe (~26%) was polarized up to 52% using the Xemed xenon polarizer and dispensed into a TedLar bag. A TedLar bag and a saline tube were connected to the membrane contactor (3M Liqui-Cel MM-0.5 × 1 Series) to mix HP Xe with saline. The outlet of the membrane contactor was connected to the 3.175 mm tube and the end of the tube was placed into the custom-built quadrature coil tuned to the Xe resonance frequency at 3T (35.33 MHz). A saline flow rate was set to 5 mL/min, 6 mL/min, 7 mL/min, and 10 mL/min.   129 Xe TOF recovery curves were measured using the TOF imaging approach. A difference in flow rate of 1 mL/min created well-detected differences in the TOF curves. A quasilinear region was observed between 200 ms and 700 ms. A nonlinear recovery appeared for longer recovery times, followed by a steady state. (c) Pearson's linear correlation between the flow rate and the slopes of the TOF curves was calculated at the quasilinear region. A strong positive correlation (r ≈ 0.988) was observed, indicating the potential to use this slope for flow mapping.

HP 129 Xe TOF Cerebral Perfusion Imaging
Human imaging experiments were approved by the research ethics boards at Lakehead University and the Thunder Bay Regional Health Science Centre and conducted in accordance with the Tri-Council Policy Statement 2. All participants signed an informed consent form. During human  129 Xe TOF recovery curves were measured using the TOF imaging approach. A difference in flow rate of 1 mL/min created well-detected differences in the TOF curves. A quasilinear region was observed between 200 ms and 700 ms. A nonlinear recovery appeared for longer recovery times, followed by a steady state. (c) Pearson's linear correlation between the flow rate and the slopes of the TOF curves was calculated at the quasilinear region. A strong positive correlation (r ≈ 0.988) was observed, indicating the potential to use this slope for flow mapping. The 129 Xe TOF GRE imaging was conducted using 20 TOF recovery times that varied from 200 ms to 2000 ms, with a step of 100 ms. The following parameters were used for imaging: TR/TE = 104.16 ms/1.22 ms, FA = 12.5 • , FOV = 100 × 100 mm 2 .

In Vivo Cerebral Perfusion Measurements in Healthy Volunteers
This research study was approved (Jan. 28, 2019) by the research ethics boards (REB) at Lakehead University (LU) and the Thunder Bay Regional Health Sciences Centre (TBRHSC) (REB file number RP-307) and conducted in accordance with the Tri-Council Policy Statement-2 (TCPS-2). Nine healthy volunteers between the ages of 21 and 77 were recruited from the community for this study. All participants signed an informed consent form. All participants were cognitively normal and consented to the data obtained from them being disseminated in this report.

1 H Magnetic Resonance Imaging
Participants were placed into a dual tuned 1 H/ 129 Xe head coil in a Philips Achieva 3T clinical MRI scanner. T2-weighted 1 H MRI were acquired using a turbo-spin echo (TSE) sequence with a Cartesian readout. The high-resolution anatomical proton images were acquired in axial and sagittal views. The sagittal images were acquired using a field-of-view equal to 250 × 250 × 46 mm 3 and a voxel size of 0.98 × 1.04 × 6 mm 3 . The repetition time was set to 3 s and the echo time was equal to 80 s. The flip angle of excitation radiofrequency pulse was equal to 90 • . Five slices were acquired, separated by 6 mm gaps. The axial scans were acquired using similar parameters. The only difference was in voxel size, which was equal to 0.98 × 0.99 × 6 mm 3 in the axial scans.

129 Xe Perfusion Mapping Using Time of Flight (TOF)
Enriched 129 Xe (83%) was polarized to~50% using a Xemed xenon polarizer and dispensed into 1 L Tedlar bags. The participant breathed in the gas and held their breath for 20 s. The dynamic 129 Xe TOF imaging pulse sequence with a Cartesian gradient echo readout was initiated simultaneously with a breath-hold. Three dynamic projection brain images were acquired during the breath-hold. The TOF recovery delays were equal to 2.5 s, 6.7 s, and 7.1 s. The field of view of gradient echo pulse sequence parameters was equal to 250 × 250 mm 2 and the acquisition matrix was set to 20 × 20. The excitation pulse flip angle was set to 20 • . The repetition time was equal to 4 ms, the eco time was equal to 0.71 ms, and the bandwidth per pixel was 382 Hz.
To conduct hemodynamic response mapping, each participant was given two Xe MRI scans, one while staring at a gray screen (baseline) and another while exposed to an external stimulus, such as a flashing visual pattern (the visual stimulus used in this study can be downloaded from the link presented in the Supplementary Material S3) or s rapid left fist squeeze (motor stimulus). Exposure to the visual stimulus started approximately 7 s prior to image acquisition. Immediately prior to image acquisition, the participant inhaled 1 L of the HP 129 Xe gas and held their breath for a period of 20 s.
A dynamic 129 Xe TOF imaging pulse sequence with a Cartesian gradient echo readout was initiated simultaneously with a breath-hold using the following parameters: FOV = 250 × 250 mm 2 , TR/TE = 4 ms/0.71 ms, FA = 20 • , BW = 382 Hz, acquisition matrix = 20 × 20. Three dynamic projection brain images were acquired during the breath-hold. The TOF recovery delays were equal to 2.5 s, 6.7 s, and 7.1 s for axial projections and 1 s, 6.5 s, and 7.1 s for the sagittal view. 129 Xe TOF perfusion-weighted images were acquired in aial view with a slice thickness of 70 mm and in sagittal view with a slice thickness of 300 mm.

129 Xe Perfusion Mapping Image Processing
Briefly, 129 Xe images were zero-padded in k-space to a 32 × 32 matrix and a Fast Fourier transform (FFT) was applied to create MR images. SNR maps were created by dividing each pixel by the standard deviation of the noise region. Based on Killian's model of dissolved Xe 18 , the mathematical Diagnostics 2020, 10, 630 5 of 13 theory was generalized and applied for HP 129 Xe perfusion mapping (for a detailed explanation, see Supplementary Material S1). Three SNR maps created from TOF images and acquired with different recovery times were fitted pixel-by-pixel using the linear equation of the SNR delay time dependence (see Supplementary Material S2), and slope maps were created. The slope maps were further recalculated into perfusion maps using the developed model (see Supplementary Material S1). The SNR of the perfusion images was calculated as the ratio of the mean pixel intensity value from the brain region to the standard deviation of the selected square region from the background. The measured SNR values were compared to the average ASL SNR at 3T published in [39] using one-sample t-test. Xe hemodynamic response maps were created by subtracting the baseline slope map from the task slope map. An Xe functional brain map was overlaid on top of the high-resolution, T 2 -weighted proton image. Additional image processing information can be found in the Supplementary Material S4.

BOLD fMRI Image Acquisition
Participants were placed into an eight-channel Philips SENSE coil tuned to the 1 H nucleus. 180 dynamic multi-slice echo-planar imaging (EPI) scans were acquired using the following parameters: FOV = 250 × 250 × 119 mm 3 , acquisition matrix = 64 × 64, voxel size = 3.91 × 3.91 × 4 mm 3 , TR/TE = 2 s/30 ms, FA = 90 • , 24 slices with a 1 mm slice gap. Participants were subjected to the stimuli described above for the first 20 s and then a rest period for 20 s. Overall, 18 stimulus/rest repetitions were presented during the BOLD fMRI scans.

BOLD fMRI Image Processing
BOLD fMRI and T 2 -weighted anatomical images were converted from Philips PAR into Analyze format, and multivolume images were segmented into multiple 3D hdr files using MRIcro 1.40 (by Chris Rorden). The first 20 EPI scans (first complete cycle) were discarded to avoid T 1 effects. fMRI data were processed using SPM12 [40] software using MATLAB R2018b (The Mathworks, Inc, Natick, MA, USA). Following manual alignment of the structural T 2 image to the average canonical T 2 image, the EPI image was manually aligned to the processed structural image. Using the SPM12 software, the obtained aligned EPI images were realigned and only the mean image was resliced. After slice timing and co-registration (an estimate), segmentation of the anatomical T 2 image was performed and the final image was corrected to remove the spatially varying artifact (modulation of the image signal intensity). The functional and anatomical images were normalized using a saved deformation field with specification of the voxel size. Smoothing of the functional images was done with Full Width at Half Maximum (FWHM) set to 6 mm in all directions. The positive t-contrast (stimulus > rest) was calculated for the final functional image. The p-value was adjusted to 0.05 with a Family-Wise Error (FWE) correction. No masking or thresholding were applied during the image processing. Statistical maps were overlaid on volume-rendered brain images provided by SPM12.
Signal enhancement was estimated using the MarsBar extension for SPM12 (by Matthew Brett, available at http://marsbar.sourceforge.net/). The region of interest (ROI) was specified for the cluster with the highest intensity from SPM results file. This was viewed and exported as an image with the base space.

In Vitro Evaluation of HP 129 Xe TOF Pulse Sequence
The 129 Xe TOF pulse sequence (Figure 1) was programmed on a Philips 3.0T Achieva MRI scanner and tested using a flow phantom (Figure 2a). A previously developed model of Xe uptake in the brain [41] was modified to facilitate quantitative perfusion measurements using 129 Xe TOF imaging. The complete analytical function, which describes the signal evolution in a brain voxel (see Supplementary Material S1), is difficult to implement practically and, therefore, a simplified model for the 129 Xe wash-in phase was developed by employing several assumptions. First, 129 Xe relaxation in blood [32,[41][42][43][44][45] is the predominant factor of polarization decay during the wash-in phase. Second, the value of the sum of the tissue perfusion-to-partition coefficient ratio plus the relaxivity of HP 129 Xe in the tissue is small. With these two assumptions, the SNR evolution during the TOF imaging sequence can be expressed as a linear function (for the detailed mathematical derivation please see Supplementary Information), where the slope of the line is directly proportional to the sum of perfusion of all tissues in the voxel. This approach is valid only for short recovery times and can be easily used in practice, albeit yielding an underestimation of the slope of the TOF recovery curves for longer TOF delay times.
The flow rates were set to 5 mL/min, 6 mL/min, 7 mL/min, and 10 mL/min. The signal-to-noise ratio (SNR) increased linearly at small recovery times, followed by a nonlinear transition (intermediate recovery times) up until saturation (long recovery times) (Figure 2b). The nonlinear dynamic at intermediate recovery times can be explained by a nonlinear 129 Xe velocity distribution in saline flow cross-sections. The signal became saturated earlier for the fastest flow rate, whereas a slight signal saturation was observed for the slowest flow rate. The slopes of the TOF recovery curves were calculated, and a strong Pearson's correlation (r = 0.988, p < 0.05) between the flow rate and the signal recovery rate was observed (Figure 2c). A difference of 1 mL/min in flow caused the observed change in the recovery rate.

HP 129 Xe TOF Cerebral Perfusion Imaging
Human imaging experiments were approved by the research ethics boards at Lakehead University and the Thunder Bay Regional Health Science Centre and conducted in accordance with the Tri-Council Policy Statement 2. All participants signed an informed consent form. During human imaging, the changes in the volunteers' blood oxygen saturation levels were within the normal range. All images were acquired during a single breath-hold after inhalation of 1 L of HP 129 Xe. The obtained images were converted from radiological to anatomical views. Figure 3a,f demonstrate the high-resolution, T 2 -weighted proton images which were used for brain localization. Figure 3b-d show 129 Xe TOF images acquired in the axial view after TOF of 2.5, 6.5, and 7.1 s. HP 129 Xe TOF sagittal images (Figure 3g-i) were acquired following 1 s, 6.5 s, and 8 s TOF delays. Following pixel-by-pixel calculations of the TOF slope, the corresponding 129 Xe TOF perfusion-weighted images were reconstructed in both the axial and sagittal projections. The average SNR of the reconstructed, perfusion-weighted images was equal to 11.2 ± 2.9 (in the sagittal projection) and 9.5 ± 2.9 (in the axial projection). Using the theoretical calculations shown in the Supplementary Information, the slope maps were transformed into net perfusion maps (Figure 3e,j). Although the 129 Xe TOF pulse sequence is similar to the ASL pulse sequence, the manipulation of the magnetization is fundamentally different. The contrast of perfusion-weighted ASL images comes from the difference between control and spin-tagged images, whereas the contrast of TOF perfusion maps comes directly from the velocity of incoming blood flow.

HP 129 Xe Hemodynamic Response Detection
The blood-flow changes due to the hemodynamic response (HDR) to visual and motor stimuli (see Supplementary Material S3) ertr detected using 129 Xe TOF perfusion imaging. The experimental designs for HP 129 Xe TOF and 1 H BOLD fMRI are shown in Figure 4a,c and Figure 5a,c. The reconstructed HDR maps nicely correlated with 1 H BOLD fMRI and demonstrated the activation of the same brain regions. Activation areas demonstrated on 129 Xe HDR maps agreed with previously published results [47][48][49][50]. Figure 4 illustrates the results obtained after subjecting the volunteer to the colorful rotating dotted visual stimulation. HP 129 Xe TOF detected activation of the visual cortex (occipital lobe), superior parietal lobe, and frontal gyrus (Figure 4b). These results completely agreed with 1 H BOLD fMRI images acquired by 129 Xe TOF image (Figure 4d).  The gradual signal-to-noise ratio increase can be observed with increasing wash-in time. The slope map was created by a pixel-by-pixel linear fit of the 129 Xe brain images. (e) The perfusion map (measured in mL of blood per mL of tissue per min) created by the pixel-by-pixel recalculation of the TOF slope was used to calculate the sum of the perfusion rates of gray and white matter superimposed on top of a high-resolution proton brain image. The calculated values of perfusion agreed with previously observed values [41,46]. (g-i) Three dynamic TOF images acquired after 1s, 6.5 s, and 8 s TOF in the sagittal view. (j) Perfusion map in the sagittal view. Similar to e, the intensity values were the net sum of the white and gray matter perfusion rates. (a,f) High-resolution, T2-weighted 1 H scans for brain localization. (b-d) Three dynamic HP 129 Xe TOF images acquired 2.5 s, 6.8 s, and 7.1 s after the application of a depolarization radiofrequency pulse in the axial projection. The image artifact in the top left corner in b is from excess gaseous 129 Xe at the end of the inhalation tube connected to the TedLar bag. The gradual signal-to-noise ratio increase can be observed with increasing wash-in time. The slope map was created by a pixel-by-pixel linear fit of the 129 Xe brain images. (e) The perfusion map (measured in mL of blood per mL of tissue per min) created by the pixel-by-pixel recalculation of the TOF slope was used to calculate the sum of the perfusion rates of gray and white matter superimposed on top of a high-resolution proton brain image. The calculated values of perfusion agreed with previously observed values [41,46]. (g-i) Three dynamic TOF images acquired after 1s, 6.5 s, and 8 s TOF in the sagittal view. (j) Perfusion map in the sagittal view. Similar to e, the intensity values were the net sum of the white and gray matter perfusion rates.

HP 129 Xe Hemodynamic Response Detection
The blood-flow changes due to the hemodynamic response (HDR) to visual and motor stimuli (see Supplementary Material S3) ertr detected using 129 Xe TOF perfusion imaging. The experimental designs for HP 129 Xe TOF and 1 H BOLD fMRI are shown in Figure 4a,c and Figure 5a,c. The reconstructed HDR maps nicely correlated with 1 H BOLD fMRI and demonstrated the activation of the same brain regions. Activation areas demonstrated on 129 Xe HDR maps agreed with previously published results [47][48][49][50]. Figure 4 illustrates the results obtained after subjecting the volunteer to the colorful rotating dotted visual stimulation. HP 129 Xe TOF detected activation of the visual cortex (occipital lobe), superior parietal lobe, and frontal gyrus (Figure 4b). These results completely agreed with 1 H BOLD fMRI images acquired by 129 Xe TOF image (Figure 4d).

Discussion
There is no HP MRI approach currently available which is suitable for quantitative perfusion imaging. Recently, HP MRI was used to detect cerebral perfusion changes caused by stroke [35] and Alzheimer's disease [34]. Since HP 129 Xe freely dissolves in the blood and travels to highly perfused organs, it can be used as a contrast agent for quantitative imaging of perfusion and blood flow. Therefore, HP 129 Xe TOF blood flow mapping is a possible next step in blood flow and perfusion measurement.
Previous works focused on the dynamic acquisition of multiple brain projection images during a breath-hold and the following wash-out time of 129 Xe from the brain [6,33,34]. Acquired images were subsequently averaged [6,33] or postprocessed to characterize the wash-out process of 129 Xe from the brain [34]. These approaches were strongly dependent on variations in the breath-hold period and the individual lung capacities of the subjects. There is typically some signal variation in the 129 Xe images acquired from the same subject from different breath-holds due to individual variation in the amount of inhaled gas, the amount of gas actually in the alveolar lung space, and the

Discussion
There is no HP MRI approach currently available which is suitable for quantitative perfusion imaging. Recently, HP MRI was used to detect cerebral perfusion changes caused by stroke [35] and Alzheimer's disease [34]. Since HP 129 Xe freely dissolves in the blood and travels to highly perfused organs, it can be used as a contrast agent for quantitative imaging of perfusion and blood flow. Therefore, HP 129 Xe TOF blood flow mapping is a possible next step in blood flow and perfusion measurement.
Previous works focused on the dynamic acquisition of multiple brain projection images during a breath-hold and the following wash-out time of 129 Xe from the brain [6,33,34]. Acquired images were subsequently averaged [6,33] or postprocessed to characterize the wash-out process of 129 Xe from the brain [34]. These approaches were strongly dependent on variations in the breath-hold period and the individual lung capacities of the subjects. There is typically some signal variation in the 129 Xe images acquired from the same subject from different breath-holds due to individual variation in the amount of inhaled gas, the amount of gas actually in the alveolar lung space, and the amount of gas remaining in the trachea and larynx. Therefore, a proper comparison of images acquired from different breath-holds is challenging and requires proper signal normalization.
The HP 129 Xe TOF technique is the first differential HP MRI technique, to our knowledge, that relies on 129 Xe signal recovery rate calculation. This imaging approach is fundamentally different from all Diagnostics 2020, 10, 630 9 of 13 other MRI methods for blood flow mapping and relies on the special properties of HP MRI. An initial depolarization radiofrequency pulse applied to the imaging region provides the necessary reference for accurate imaging and makes the obtained perfusion-weighted images independent of individual lung capacity and breath-hold variations. In addition, the HP 129 Xe TOF approach allows quantitative perfusion mapping using the derived theory (see Supplementary Material S1).
Although HP 129 Xe blood flow mapping is similar to ASL in terms of pulse sequences, this technique overcomes the issue of low signal-to-noise ratio, which is typical for ALS perfusion images (the mean SNR of the Xe perfusion maps was approximately double compared to values acquired for ASL [39] (p < 0.05)) [1,51]. The obtained 129 Xe perfusion maps (Figure 3e,j) looked similar to results previously obtained from healthy volunteers acquired using ASL MRI [52]. With the additional separation of the HP 129 Xe signals originating from various brain tissues from the signal of 129 Xe dissolved in blood (either via different phase separation methods or using chemical shift selective imaging), it should be possible to quantitively map the perfusion of each tissue in a region of interest. Signal separation of the dissolved 129 Xe phases may allow the use of an exact mathematical solution for perfusion map calculations (see Supplementary Material S1), which would eliminate any potential underestimation of perfusion caused by using an approximate linear model.
Recently, multiple advances in ASL perfusion imaging were achieved by developing a pseudocontinuous ASL (pCASL), resulting in an SNR increase compared to conventional ASL [16][17][18]53]. pCASL perfusion images have higher spatial resolution compared to the present HP 129 Xe TOF images, mainly due to their significantly better developed hardware (multichannel coils used for parallel imaging during perfusion imaging) and software available for conducting proton ASL MRI. By using a parallel imaging approach, compressed sensing, nonCartesian k-space trajectories, and by developing multichannel 129 Xe coils, it should be possible to acquire HP 129 Xe perfusion maps with a spatial resolution comparable to pCASL resolution and with a significantly higher SNR.
Although the scan time for acquiring a single brain perfusion map was~20 s, it could be shortened to~10 s by reducing TOF recovery time delays. The scan time could be shortened further for kidney perfusion measurements, since the reasonably high SNR of HP 129 Xe dissolved in the kidneys can be acquired in less than 4 s [35].
As one practical application, the HP 129 Xe perfusion mapping technique could be used for functional imaging of the human brain, which would allow direct hemodynamic response mapping. HP 129 Xe functional brain maps can be acquired significantly faster compared to traditional BOLD functional images. In addition, this approach requires less postprocessing than the BOLD technique, which makes it less ambiguous in terms of data processing. Although BOLD fMRI and HP 129 Xe perfusion mapping are two fundamentally different techniques and quantitative comparisons between them are challenging, the estimated percent signal enhancement of the functional maps obtained by HP 129 Xe perfusion mapping was up to two orders of magnitude larger than the BOLD fMRI images. Xenon gas is used in clinics as a general inhalation anesthetic with continuous breathing over several minutes at a total lung concentration greater than 50% [54]. Previous studies demonstrated an acceptably low incidence of adverse effects caused by inhalation of xenon with an overall alveolar concentration below 32% [55]. The concentration of inhaled HP 129 Xe in our study was 20-25% (considering a lung volume of 4-5 L), and there were no side effects observed after multiple inhalations and breath-holds.
A major limitation of the approach demonstrated for 129 Xe TOF functional imaging is the requirement of acquiring projection images instead of multislice images due to the limited quantity of HP 129 Xe that is transported to the brain. This issue can be overcome by using parallel imaging and compressed sensing techniques, which allow for faster acquisition with higher SNR per TOF image. Therefore, a larger number of slices can be scanned during a single breath-hold. Another limitation of HP 129 Xe perfusion mapping, in terms of its clinical relevance, is the ability of a participant to hold their breath. Although the short scan time required allows perfusion mapping in patients with lung diseases who can only tolerate short breath-hold periods, this technique is not possible for participants who cannot hold their breath for at least eight seconds. As an alternative, continuous breathing of an Diagnostics 2020, 10, 630 10 of 13 HP 129 Xe mixture with oxygen could be considered. A continuous breathing protocol has so far only been utilized for animal lung studies [56,57]. Translation of this breathing protocol to HP perfusion imaging might be challenging due to the oxygen-induced T 1 shortening of HP 129 Xe and subsequent decrease in the 129 Xe MRI signal. On the other hand, this approach could allow for multiple signal averaging and expand the application of HP 129 Xe TOF perfusion imaging. As a particular example, patients who are not able to hold their breath for eight seconds could be scanned using continuous breathing. Moreover, this might allow for HP 129 Xe perfusion imaging during clinical procedures when patients are under HP 129 Xe-induced anesthesia. Since the 129 Xe TOF technique is based on selective depolarization of 129 Xe nuclei, the MRI scanner needs to be properly calibrated, otherwise there will be an additional contribution to the signal from partially depolarized nuclei. Another potential limitation for some HP perfusion imaging applications is the time needed for 129 Xe polarization. The polarizer used in this study requires about 15 min to polarize 1 L of 129 Xe. Therefore, as a diagnostic tool, HP 129 Xe TOF might not be suitable for urgent imaging purposes. On the other hand, Xe was shown to demonstrate neuroprotective properties [58], indicating a potential use of this modality for cerebral perfusion imaging in stroke patients, for example. Finally, the other limitations of this technique are the high cost of enriched 129 Xe gas and the polarizer instrument. It should be noted that the use of enriched 129 Xe is required by an overall low HP 129 Xe signal in the brain. With further hardware development, the use of naturally abundant 129 Xe, which is substantially less expensive, is expected to become more feasible.
Overall, HP 129 Xe perfusion mapping is a novel HP MRI method which, we believe, could expand future knowledge in the field of perfusion imaging and functional brain mapping. It may also be a useful tool for the detection of perfusion changes in different organs and for further diagnostics regarding perfusion-related diseases.

Patents
Based on the results of this work, the provisional patent entitled "Method to Detect Brain Functional Activities Using Hyperpolarized 129 Xe MR" was filed.